A Framework for Evaluating Early-Stage Human - of Marcus Hutter
A Framework for Evaluating Early-Stage Human - of Marcus Hutter
A Framework for Evaluating Early-Stage Human - of Marcus Hutter
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
The China-Brain Project<br />
Report on the First Six Months<br />
Pr<strong>of</strong>. Dr. Hugo de GARIS, Dean Pr<strong>of</strong>. Dr. ZHOU Changle, Pr<strong>of</strong>. Dr. SHI Xiaodong,<br />
Dr. Ben GOERTZEL, Pr<strong>of</strong>. PAN Wei, Pr<strong>of</strong>. MIAO Kehua,<br />
Pr<strong>of</strong>. Dr. ZHOU Jianyang, Dr. JIANG Min, Pr<strong>of</strong>. ZHEN Lingxiang,<br />
Pr<strong>of</strong>. Dr. WU Qinfang, Pr<strong>of</strong>. Dr. SHI Minghui, LIAN Ruiting, CHEN Ying<br />
Abstract<br />
The “China Brain Project” is a 4 year (2008-2011),<br />
10.5 million RMB research project to build China’s<br />
first artificial brain, which will consist <strong>of</strong> 10,000-<br />
50,000 neural net modules which are evolved rapidly<br />
in special FPGA hardware, downloaded one by one<br />
into a PC or supercomputer, and then connected<br />
according to the designs <strong>of</strong> human “BAs” (brain<br />
architects) to build an artificial brain with thousands<br />
<strong>of</strong> pattern detectors to control the hundreds <strong>of</strong><br />
behaviors <strong>of</strong> a two legged robot.<br />
1. Introduction<br />
The “China Brain Project”, based at Xiamen<br />
University, is a 4 year (2008-2011), 10.5 million<br />
RMB, 20 person, research project to design and<br />
build China’s first artificial brain (AB). An<br />
artificial brain is defined here to be a “network <strong>of</strong><br />
(evolved neural) networks”, where each neural<br />
net(work) module per<strong>for</strong>ms some simple task<br />
(e.g. recognizes someone’s face, lifts an arm <strong>of</strong><br />
a robot, etc), somewhat similar to Minsky’s idea<br />
<strong>of</strong> a “society <strong>of</strong> mind” [1], i.e. where large<br />
numbers <strong>of</strong> unintelligent “agents” link up to<br />
create an intelligent “society <strong>of</strong> agents”. 10,000s<br />
<strong>of</strong> these neural net modules are evolved rapidly,<br />
one at a time, in special (FPGA based) hardware<br />
and then downloaded into a PC (or more<br />
probably, a supercomputer PC cluster). <strong>Human</strong><br />
“BAs” (brain architects) then connect these<br />
evolved modules according to their human<br />
designs to architect artificial brains. Special<br />
Artificial Brain Lab<br />
Cognitive Science Department<br />
School <strong>of</strong> In<strong>for</strong>mation Science and Technology<br />
Xiamen University, Xiamen, China<br />
pr<strong>of</strong>hugodegaris@yahoo.com<br />
25<br />
s<strong>of</strong>tware, called IMSI (see section 5) is used to<br />
specify these connections, and to per<strong>for</strong>m the<br />
neural signaling <strong>of</strong> the whole brain (in real time).<br />
The main aim <strong>of</strong> this research project is to show<br />
that using this (evolutionary engineering)<br />
approach to brain building is realistic, by simply<br />
building one and show that it can have thousands<br />
<strong>of</strong> pattern recognizers, and hundreds <strong>of</strong> motions<br />
that are switched between, depending on external<br />
and internal stimuli. This project already has<br />
(Fujian) province “key lab” financial support. It<br />
is hoped, in three years, that it will win “key<br />
state (i.e. federal) lab” status. In 5-10 years, it is<br />
hoped that China will establish a “CABA”<br />
(Chinese Artificial Brain Administration),<br />
consisting <strong>of</strong> thousands <strong>of</strong> scientists and<br />
engineers, to build national brains to fuel the<br />
home robot industry (which may become the<br />
worlds largest) (See section 11.)<br />
There are about 20 people (8 pr<strong>of</strong>essors)<br />
involved in this project, divided into specialist<br />
teams, i.e.<br />
a) “Vision team” (who evolve pattern<br />
recognition modules and create vision system<br />
architectures).<br />
b) ”Robot team” (who program the NAO robot<br />
[2] to per<strong>for</strong>m the many (hundreds <strong>of</strong>) behaviors<br />
that the robot is to per<strong>for</strong>m).<br />
c) “Hardware Acceleration team” (who<br />
program the FPGA electronic boards we use to